ABSTRACT
With the coming of the era of ‘big data’, people are able to collect data easily. However, most of the data are not labeled and it wastes time and energy to label them. To find hidden structure and useful information from unlabeled data, unsupervised learning was proposed, where there is no supervision to guide the learning process. In this note, I will briefly introduce some classical unsupervised learning methods, mainly the clustering methods, and state the challenges caused by big data, in terms of both efficiency and effectiveness. Some most recent methods for large-scale unsupervised learning techniques, and the corresponding applications in regional science, will be stated as well.
